Classification of poor contrast mammograms using a novel and fast boundary detection technique

It is well established that edges are boundaries between different intensities and therefore preserve important structural properties of the image. Therefore efficient edge detection mechanism is important in analysis of poor contrast mammographic images. In this study we have presented a computationally efficient and fast procedure to estimate the value of global threshold by which the object boundary of images characterized by bimodal histograms will be identified. The proposed algorithm is shown to be relatively insensitive to gradually varying interference. Moreover, the memory space required is independent of image size which also makes the implementation of proposed approach very cost effective. After detecting the mass boundaries, multilayer neural network is used as classifier for discrimination of benign/malignant masses. The results show that the proposed method can successfully segment the mammograms and provide improved classification rate than Canny segmentation approach.